MNISTを用いた学習モデルの作成における，KeyErrorが解消できません．

前提

開発環境はmacos，spyderです

発生している問題・エラーメッセージ

File ~/anaconda3/lib/python3.10/site-packages/pandas/core/indexes/base.py:6130 in _raise_if_missing raise KeyError(f"None of [{key}] are in the [{axis_name}]") KeyError: "None of [Int64Index([35591, 20294, 18337, 16931, 13072, 69126, 39634, 25769, 11346,\n 51243, 21720, 20385, 62344, 43301, 26070, 34432, 34125, 68804,\n 51753, 36023, 69494, 28196, 9979, 6770, 32787, 41588, 68411,\n 11848, 25625, 7457, 28528, 24608, 9720, 58788, 29144, 64967,\n 24248, 3573, 20452, 40200, 789, 69971, 61182, 64594, 21509,\n 69578, 64162, 24641, 22162, 29223, 57707, 16098, 27149, 37215,\n 29198, 53891, 26750, 28734, 60063, 21480, 40913, 39801, 65790,\n 50183, 57479, 17439, 793, 35548, 15707, 3294, 62658, 44846,\n 3386, 7345, 53687, 19329, 69516, 52891, 1857, 24471, 49636,\n 62473, 65023, 66143, 8098, 59403, 36681, 60229, 34429, 53443,\n 38653, 13604, 53509, 13794, 67996, 11017, 51464, 35539, 69727,\n 13377],\n dtype='int64')] are in the [columns]"

試したこと

おそらくこの部分がおかしいのではないかと思い，.valuesを加えたりと試したのですが解消されませんでした

for i in range(0, DATASIZE, BATCHSIZE): x = Variable(x_all[indexes[i : i + BATCHSIZE]]) t = Variable(y_all[indexes[i : i + BATCHSIZE]])

該当のソースコード

python

1import numpy as np 2from chainer import Variable, optimizers, serializers 3from chainer import Chain 4import chainer.functions as F 5import chainer.links as L 6from sklearn.datasets import fetch_openml 7 8 9class MyMLP(Chain):10 def __init__(self, n_in=784, n_units=100, n_out=10):11 super(MyMLP, self).__init__(12 l1=L.Linear(n_in, n_units),13 l2=L.Linear(n_units, n_units),14 l3=L.Linear(n_units, n_out),15 )16 def __call__(self, x):17 h1 = F.relu(self.l1(x))18 h2 = F.relu(self.l2(h1))19 y = self.l3(h2)20 return y 21 22print('Start')23 24mnist_X, mnist_y = fetch_openml('mnist_784', version=1, data_home=".", return_X_y=True)25 26x_all = mnist_X.astype(np.float32) / 25527y_all = mnist_y.astype(np.int32)28 29model = MyMLP()30optimizer = optimizers.SGD()31optimizer.setup(model)32 33BATCHSIZE = 10034DATASIZE = 7000035 36for epoch in range(20):37 print('epoch %d' % epoch)38 indexes = np.random.permutation(DATASIZE)39 for i in range(0, DATASIZE, BATCHSIZE):40 x = Variable(x_all[indexes[i : i + BATCHSIZE]])41 t = Variable(y_all[indexes[i : i + BATCHSIZE]])42 43 model.zerograds()44 45 y = model(x)46 47 loss = F.softmax_cross_entropy(y, t)48 49 loss.backward()50 51 optimizer.update()52 53serializers.save_npz("mymodel.npz", model)54 55print('Finish')